** Apologies if you receive multiple copies of this
announcement **

*******************************************************************
Second Call for Papers

2nd International Conference on Hybrid Intelligent
Systems (HIS'02)

December 01 - 04, 2002
Universidad de Chile, Santiago, Chile

http://his02.hybridsystem.com

HIS’02 is technically co-sponsored by:
- IEEE Region 9
- The World Federation on Soft Computing
- European Neural Network Society
- European Society for Fuzzy Logic and Technology
- European Network of Excellence in Evolutionary
Computing
- IOS Press.

********************************************************************
Deadline for Paper Submission: July 31, 2002
********************************************************************

HIS'02 is the second International conference that
brings together
researchers, developers, practitioners, and users of
soft computing,
computational intelligence, agents, logic programming,
and several
other intelligent computing techniques. The aim of
HIS'02 is to serve
as a forum to present current and future work as well
as to exchange
research ideas in this field.

HIS'02 invites authors to submit their original and
unpublished work
that demonstrate current research using soft
computing/computational
intelligence and other intelligent computing
techniques and their
applications in science, technology, business and
commercial. Please
submit a full paper of 8 to 10 pages (Letter or A4
paper) for oral
presentation. The proceedings of the Conference will
be published by
IOS Press, Netherlands. A selected number of papers
will also be
considered for a special issue of the Elsevier Science
Journal
"Applied Soft Computing" and for a special issue of
the IOS Press
Journal- Intelligent Data Analysis. Please follow the
author’s
guidelines given by IOS Press for more information on
submission.
Author's guidelines can be downloaded from the
conference web page.

*******************************************************************
Topics of Interest include but are not limited to:

Track 1: Computational Intelligence and Applications

- Artificial neural networks and learning techniques
- Artificial neural network optimization using global
optimization
   techniques
- Interactions between neural networks and fuzzy
inference systems
- Fuzzy clustering algorithms
- Fuzzy system optimization using global optimization
algorithms
- Evolutionary computation
- Support vector machines, rough sets, Bayesian
networks
- Hybrid computing using neural networks-fuzzy
systems- evolutionary
   algorithms
- Hybrid optimization techniques
- Hybrid of soft computing and statistical learning
techniques
- Intelligent agents
- Models using inductive logic programming, logic
synthesis
- Natural computation techniques
- Autonomic computing
- Applications using the above systems
- Special topics


Track 2: Soft Computing for Pattern Recognition and
Signal Processing

- Features and Classification
- Texture Analysis
- Document Analysis
- Shape Processing
- Fuzzy Image Processing
- Grouping and Segmentation
- Object Recognition
- Medical Image Processing
- Image and Video Retrieval
- Biometric Systems
- Image Representation
- Image Compression
- Video Processing
- Digital Watermarking
- Image Synthesis
- Industrial Applications
- Visual Surveillance


Track 3: Soft Computing and the Internet

- e-learning, e-commerce, e-business
- Web Intelligence
- Search Engines
- Information retrieval (web mining)
- Database Querying
- Ontology
- XML mining
- Intelligent networking between Web Sites
- Content Management
- Information Aggregation and Fusion
- Interaction with Intelligent Agents
- Intelligent Agents and Interfaces for
personalization and adaptivity
- Intelligent tutoring systems on the WWW
- WWW Recommender Systems
- Adaptive Hypermedia Systems
- Agents for Digital cities, virtual communities and
agent societies


Track 4: Intelligent Data Mining

- Integrating E-Commerce and Data Mining
- Discovering patterns in Continuous Data
- Uncertainty management for data mining
- Clustering algorithms and applications
- Classification trees
- Mining time series
- Mining in a Mobile Environment
- Statistical Considerations in Learning
- XML Mining
- Text Mining
- Distributed Data Mining


*******************************************************************
HIS02 Important Dates
*******************************************************************

Deadline for Events Proposals                    June
30, 2002
Deadline for Paper Submission (full paper)       July
31, 2002
Notification of Acceptance
August 30, 2002
Deadline for Camera Ready Papers
September 10, 2002
HIS'02 Conference in Chile
December 01-04, 2002

*******************************************************************
Venue Information (Santiago, Chile)

HIS’02 will be hosted by the Faculty of Physical and
Mathematical
Sciences at the Universidad de Chile (Beaucheff 850,
Santiago
de Chile).

December is an excellent time to travel in Chile. The
local
organizers will help the interested participants to
arrange a
trip to any location in Chile. For more information
about traveling
in Chile please visit:
http://www.sernatur.cl/ and
http://www.visit-chile.org/

********************************************************************

HIS02 Organization

Honorary Chairman
Yasuhiko Dote, Muroran Institute of Technology, Japan

General Chairmen
Ajith Abraham, Monash University, Australia
Mario Köppen, Fraunhofer IPK-Berlin, Germany
Javier Ruiz-del-Solar, Universidad de Chile, Chile

Local Organizing Chair
Javier Ruiz-del-Solar, Universidad de Chile, Chile

Area Chairs
Track1: Computational Intelligence and Applications
Ajith Abraham, Monash University, Australia

Track 2: Soft Computing for Pattern Recognition
and Signal Processing
Mario Köppen, Fraunhofer IPK-Berlin, Germany

Track 3: Soft Computing and the Internet
Fabio Abbattista, Universitŕ di Bari, Italy

Track 4: Intelligent Data Mining
Kate Smith, Monash University, Australia

Web Chairs
Mauricio Gaete, Universidad de Chile, Chile
Diego Sepulveda, Universidad de Chile, Chile

Finance Chair
Roberto Aviles, Universidad de Chile, Chile

Local Organizing Committee
Roberto Aviles, Universidad de Chile, Chile
Rodrigo Palma, Universidad de Chile, Chile
Richard Weber, Universidad de Chile, Chile

International Technical Committee

Sung-Bae Cho, Yonsei University, Korea
Kate Smith, Monash University, Australia
Hung T. Nguyen, New Mexico State University, USA
Tom Gedeon, Murdoch University, Australia
Robert John, De Montfort University, UK
Fabio Abbatista, Universitat di Bari, Italy
Bernard de Baets, Ghent University, Belgium
Carlos A. Coello Coello,Nacional de Informática
Avanzada, Mexico
Vladimir Kvasnicka, Slovak Technical University,
Slovakia
Zensho Nakao, University of Ryukyus, Japan
Witold Pedrycz, University of Alberta, Canada
Udo Seiffert, University of South Australia, Australia
Jarno Tanskanen, University of Kuopio, Finland
Hamid R. Tizhoosh, University of Waterloo, Canada
Olgierd Unold, Wroclaw University of Technology,
Poland
Vasant Honavar, Iowa State University, USA
Maumita Bhattacharya, Monash University, Australia
José Manuel Benítez, University of Granada, Spain
Janusz Kacprzyk, Polish Academy of Sciences, Poland
Sankar K Pal, Indian Statistical Institute, India
Vijayan Asari, Old Dominion University, USA
Dharmendhra Sharma, University of Canberra, Australia
Morshed U Chowdhury, Deakin University, Australia
Xiao Zhi Gao, Helsinki University of Technology,
Finland
Costa Branco P J, Instituto Superior Technico,
Portugal
Zorica Nedic, University of South Australia, Australia
Rajkumar Roy, Cranfield University, United Kingdom
Katrin Franke, Fraunhofer IPK-Berlin, Germany
Aureli Soria-Frisch, Fraunhofer IPK-Berlin, Germany
Xiufen Liu, Fraunhofer IPK-Berlin, Germany
Nikhil R. Pal, Indian Statistical Institute, India
Ricardo Baeza, Universidad de Chile, Chile
Kaori Yoshida, Kyushu Institute of Technology, Japan
William B. Langdon, University College London, UK
Akira Asano, Hiroshima University, Japan
Evgenia Dimitriadou, Vienna University of Technology,
Austria
Francisco Herrera, University of Granada, Spain
Stefano Cagnoni, University of Parma, Italy
Marley Vellasco, PUC-RJ, Brazil
André Ferreira de Carvalho, Universidade de Săo Paulo,
Brazil
Antônio de Pádua Braga, Federal University of Minas
Gerais,Brazil
Etienne Kerre, Ghent University, Belgium
Frank Hoffmann, Royal Institute of Technology, Sweden
Günther Raidl, Vienna University of Technology,
Austria
Greg Huang, MIT, USA
Sami Khuri, San Jose University, USA
Hans Heinrich Bothe, Technical University of Denmark,
Denmark
Andreas König, Technical University of Dresden,
Germany
Vladislav B. Valkovsky, Uppsala University, Sweden
Luis Magdalena, Universidad Politecnica de Madrid,
Spain
L.C. Jain, University of South Australia, Australia
Jose Ramon Alvarez Sanchez, Univ. Nac.de Educacion a
Distancia, Spain
Janos Abonyi, University of Veszprem, Hungary
Hans-Michael Voigt, GFAI, Germany
Jose Mira, UNED, Spain
Andrew Sung, New Mexico Institute of Mining and
Technology,USA
Rajiv Khosla, La Trobe University, Australia



http://www.sold.com.au - SOLD.com.au
- Find yourself a bargain!

Hello,

This email message is a notification to let you know that
a file has been uploaded to the Files area of the ecomas
group.

   File        : / Click Here!
   Uploaded by : polojaeriaters <polojaeriaters@...>
   Description : Look at all these different credit cards you can get! Good, bad,
or no credit.

You can access this file at the URL

http://groups.yahoo.com/group/ecomas/files/%20Click%20Here%21

To learn more about file sharing for your group, please visit

http://help.yahoo.com/help/us/groups/files

Regards,

polojaeriaters <polojaeriaters@...>

Hello there Mark, Good to hear from a new ECOMAS member.

Re: Sim-Agent toolkit.

It is good to see philosophy feeding back into technology, although I can't
claim to be unbiased, (as a philosophy student myself). I am interested to
know how you think Evolutionary processes can be used to design agent
behaviour in such a 'cognitively rich' rule-based AI type approach as that
reflected in the sim-agent toolkit.

My work is more focused on the use of evolution to tune behaviour in
'dynamic systems' leading to cognitively-sufficient responsive agents which
can be coupled together in MAS. This includes 'rule-based' approaches such
as those found in classifier systems, but I don't go as far as the
'(physical) symbol system' in which tokens represent states or items in the
world (this is how I interpret the objectives of sim-agent). Can natural
selection and symbolic logic can be coupled together? Do you have techniques
in mind?

Re: Gaming

The results which you have in mind (creating rich and open-ended worlds for
gaming) are very interesting. I wonder whether there may be opportunities to
pursue the 'dynamic systems' type approach in this work too, allowing novel
strategies to evolve as free-living populations within the game, or
alternatively be nurtured by users and released into the game.

Making this truly open-ended requires behaviour to be constrained by a
self-consistent physics, within which a substantial and unanticipated range
of strategies can be embodied, and providing a context for the process of
natural selection. This is not easy, especially if you don't want people to
find loopholes in the physics. Mind you, how many games are truly
open-ended?

Re: Toolkit list

We should certainly update the list of toolkits on the ECOMAS site, and I
would welcome further suggestions from our subscribers if they can think of
any which aren't represented. The list is at
http://groups.yahoo.com/group/ecomas/files/agentkit.html .

Cefn

-----Original Message-----
From: Mark Gemmell [mailto:thetopmark@...]
Sent: 30 March 2002 17:27
To: ecomas@yahoogroups.com
Subject: [ecomas] Hello there


Hi everyone,

I'm a new member to ECOMAS.  My name is Mark, and I'm interesting especially

in researching ways to produce multi-agent systems that can be incorporated
in AI within computer games.

I hope to developed a sophisticated agent 'society' developed from the many
agents each with individual agendas and learning trends.  The overall hope
is to remove the linear pre-defined nature of the most plots with computer
games, and replace them with genuinely rich, realistic, unpredictable
'plots' or evolving stories.

So far my research has been exploring different agent architectures using
the agent toolkit "sim_agent" (available free of charge from
www.cs.bham.ac.uk/~axs)

While pursuing this further, I hope to also broaden my knowledge of agent
toolkits.  If any one has any experience of advise on this front, I'd
appreciate it.

I look forward to being part of the group, and sharing ideas,

cheers

Mark



_________________________________________________________________
Chat with friends online, try MSN Messenger: http://messenger.msn.com



To unsubscribe from this group, send an email to:
ecomas-unsubscribe@yahoogroups.com
---
ECOMAS Web site: http://www.csm.uwe.ac.uk/~rsmith/ECOMAS/
Discussion Group Site: http://groups.yahoo.com/group/ecomas


Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/

## Our apology, if you receive multiple copies of this message ##

IKOMAT'02,
16-18 September 2002,
Crema, Italy

Dear Colleague,

To accommodate numerous requests from interested contributors,
the submission deadline for IKOMAT'02( First International Workshop
on Intelligent Knowledge Management Techniques ) has been extended
to 15th April'02.
For further details, please refer to:

http://www.gscit.monash.edu.au/conferences/ikomat02/ikomat-02.htm

<< Important Dates >>

Submission deadline: April 15, 2002
Notification of acceptance: April 30, 2002
Camera-ready papers due: May 18, 2002

If any of you would like a special extension for submission of
papers,
please let us know as early as possible.

IKOMAT'02 Team

Hi everyone,

I'm a new member to ECOMAS.  My name is Mark, and I'm interesting especially
in researching ways to produce multi-agent systems that can be incorporated
in AI within computer games.

I hope to developed a sophisticated agent 'society' developed from the many
agents each with individual agendas and learning trends.  The overall hope
is to remove the linear pre-defined nature of the most plots with computer
games, and replace them with genuinely rich, realistic, unpredictable
'plots' or evolving stories.

So far my research has been exploring different agent architectures using
the agent toolkit "sim_agent" (available free of charge from
www.cs.bham.ac.uk/~axs)

While pursuing this further, I hope to also broaden my knowledge of agent
toolkits.  If any one has any experience of advise on this front, I'd
appreciate it.

I look forward to being part of the group, and sharing ideas,

cheers

Mark



_________________________________________________________________
Chat with friends online, try MSN Messenger: http://messenger.msn.com

## Our apology, if you receive multiple copies of this
message ##

Dear Colleague,

This is an invitation to I-KOMAT'2002: First
International Workshop on
Intelligent Knowledge Management Techniques
(I-KOMAT'2002) to be held in
conjunction with: KES'2002 (Sixth International
Conference on
Knowledge-Based Intelligent Information & Engineering
Systems, 16-18
September 2002, Podere d'Ombriano, Crema, Italy). For
detailed
information, please visit our web site:

http://www.gscit.monash.edu.au/conferences/ikomat02/ikomat-02.htm

<< Important Dates >>

Submission deadline: March 31, 2002
Notification of acceptance: April 22, 2002
Camera-ready papers due: May 18, 2002

If any of you would like a special extension for
submission of papers,
please let us know as early as possible.

IKOMAT'02 Team

http://movies.yahoo.com.au - Yahoo! Movies
- Vote for your nominees in our online Oscars pool.

Hi Cefn

My work up until now has concentrated on the simple migration of
rules between classifier systems...that is a percentage of the
population of classifiers is chosen at random (without replacement)
from each agent and passed to another neighbouring agent.  By this
approach a learning speed up and improved solution quality are seen.

Regarding the work on evolution of symbolic language, research is
still at an early stage.  I would be interested in any references or
suggestions from the group.  I have reviewed some of Luc Steels' work
and am interested in the idea of two agents being able to
induce/evolve a basic grammatical structure in a co-evolutionary
sense, i.e. they both have a fixed set of "symbols" which they
communicate to the other, and I'd like to see if a stable set of call-
response interaction can evolve.  As I say I have not begun
experimental work on this and am still at the preliminary stage of
literary research.

My main PhD focus is on Interactive Evolutionary Design and will
focus at a later stage on a different aspect of multi-agent systems.
Again, I am only at the preliminary stage of this aspect of my
research work.

Dave Wyatt


--- In ecomas@y..., cefn.hoile@b... wrote:
> Sounds like interesting work, David.
>
> What representation do you use for a symbolic language?
Specifically, how is
> it grounded in the agent's world?
>
> I can imagine how you could evolve a language in the sense of a
grammar (a
> definition of validity for the language), but I don't know how this
would be
> extended to semantics (a definition of meaning for the language).
Is this
> addressed in the work you have already done?
>
> curious,
>
> Cefn
>
> -----Original Message-----
> From: diwyatt [mailto:David2.Wyatt@u...]
> Sent: 03 March 2002 23:11
> To: ecomas@y...
> Subject: [ecomas] New member
>
>
> Hi
>
> I am a PhD research student working at UWE, Bristol as part if the
> ICSC research group.  My PhD is entitled 'LCS in Interactive
> Evolutionary Design' with supervisors Dr. L Bull and Prof I
Parmee.
> I completed a MSc degree last year and as part of the research
phase
> I investigated a multi-agent system based on ZCS classifier system
in
> which the classifiers communicated on a basic passing of induced
> rules.  I have an interest in the communication between agents esp
in
> respect of evolution of symbolic languages.
>
> Thanks
> Dave Wyatt
>
>
>
> To unsubscribe from this group, send an email to:
> ecomas-unsubscribe@y...
> ---
> ECOMAS Web site: http://www.csm.uwe.ac.uk/~rsmith/ECOMAS/
> Discussion Group Site: http://groups.yahoo.com/group/ecomas
>
>
> Your use of Yahoo! Groups is subject to
http://docs.yahoo.com/info/terms/

Sounds like interesting work, David.

What representation do you use for a symbolic language? Specifically, how is
it grounded in the agent's world?

I can imagine how you could evolve a language in the sense of a grammar (a
definition of validity for the language), but I don't know how this would be
extended to semantics (a definition of meaning for the language). Is this
addressed in the work you have already done?

curious,

Cefn

-----Original Message-----
From: diwyatt [mailto:David2.Wyatt@...]
Sent: 03 March 2002 23:11
To: ecomas@yahoogroups.com
Subject: [ecomas] New member


Hi

I am a PhD research student working at UWE, Bristol as part if the
ICSC research group.  My PhD is entitled 'LCS in Interactive
Evolutionary Design' with supervisors Dr. L Bull and Prof I Parmee.
I completed a MSc degree last year and as part of the research phase
I investigated a multi-agent system based on ZCS classifier system in
which the classifiers communicated on a basic passing of induced
rules.  I have an interest in the communication between agents esp in
respect of evolution of symbolic languages.

Thanks
Dave Wyatt



To unsubscribe from this group, send an email to:
ecomas-unsubscribe@yahoogroups.com
---
ECOMAS Web site: http://www.csm.uwe.ac.uk/~rsmith/ECOMAS/
Discussion Group Site: http://groups.yahoo.com/group/ecomas


Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/

Hi

I am a PhD research student working at UWE, Bristol as part if the
ICSC research group.  My PhD is entitled 'LCS in Interactive
Evolutionary Design' with supervisors Dr. L Bull and Prof I Parmee.
I completed a MSc degree last year and as part of the research phase
I investigated a multi-agent system based on ZCS classifier system in
which the classifiers communicated on a basic passing of induced
rules.  I have an interest in the communication between agents esp in
respect of evolution of symbolic languages.

Thanks
Dave Wyatt

*****************************************************
Our apology, if you receive multiple copies of this message
*****************************************************

Dear Colleague,

This is an invitation to I-KOMAT'2002: International Workshop on
Intelligent Knowledge Management Techniques (I-KOMAT'2002) to
be held in conjunction with: KES'2002 (Sixth International Conference
on
Knowledge-Based Intelligent Information & Engineering Systems,
16, 17 & 18 September 2002, Podere d'Ombriano, Crema, Italy)

For detailed information, please visit our web site:

http://www-mugc.cc.monash.edu.au/~maum/ikomat-02.htm



**KNOWLEDGE MANAGEMENT - THE ARTIFICIAL INTELLIGENCE
WAY**

International Workshop on Intelligent Knowledge Management
Techniques (I-KOMAT' 2002) is a technical forum for exchanging
current and future research results /ideas involving Computational
Intelligence for identification, analysis and control of knowledge
assets and processes. The scope of I-KOMAT' 2002 covers both
design and application aspects of various Artificial Intelligence
Techniques like Artificial Neural Networks, Fuzzy Systems,
Evolutionary algorithms, Hybrid systems, Agents, Support
Vector Machine etc.

I-KOMAT' 2002 offers an Interdisciplinary Platform to all
researchers, developers and practitioners dealing with AI
based modeling, ontology and planning techniques
supporting knowledge management.

Authors are invited to submit their original and unpublished
work that demonstrate current research in all areas of
computational intelligence including design of artificial
neural networks, fuzzy systems, evolutionary algorithms,
hybrid systems, agents, and their applications to knowledge
management.



Topics of interest include but not limited to:

*Design and modeling aspects of the following Computational
   Intelligence Techniques and their applications to science,
technology,
   commerce, social science and economics:

-Artificial Neural Networks and Probabilistic Reasoning
-Genetic Algorithms
-Evolution Strategies
-Genetic Programming
-Evolutionary Programming
-Learning Classifiers
-Fuzzy Systems including Fuzzy logic and possibility
theory, Fuzzy expert systems, Fuzzy system modeling
and simulation
-Hybrid Soft Computing Techniques like Neuro-Fuzzy,
fuzzy-GA, GA-Fuzzy etc.
-Intelligent Agents including Multi-Agents, Autonomous Agents
and Cooperative Agents
-Support Vector Machines
-Bayesian networks and probabilistic reasoning
-Rough sets
-Case-based Reasoning

*Data and Knowledge Representation: planning and description logics

*Data and knowledge visualization

*Knowledge acquisition

*Models and frameworks of Data Mining techniques

*Reasoning methods and technologies

*Knowledge and belief, belief revision and update, non-monotonic
formalisms, uncertainty


*****************************************************
Paper Submission
*****************************************************

Authors of contributed and invited papers are requested to submit,
before March 31, 2002, an article not exceeding five pages, of their
research presentation. Submission implies the willingness of at least
one of the authors to register and present the paper. The conference
proceedings will be published worldwide by IOS Press, Amsterdam,
the Netherlands.

Papers must correspond to the requirements detailed in IOS
Instructions
for the Preparation of a Camera-Ready Manuscript. All full papers are
to
be submitted in PDF, postscript or MS word version electronically to:
maumita.bhattacharya@....

Hard copies should be sent only if electronic submission is not
possible.
All papers will be peer reviewed by at least two independent referees
of
the international program committee of I-KOMAT'02.

Extended versions of selected papers will be considered for
publication in
the KES Journal (International Journal of Knowledge-Based Intelligent
Engineering Systems).

******************************************************
Important Dates
******************************************************
Submission deadline: March 31, 2002
Notification of acceptance: April 22, 2002
Camera-ready papers due: May 18, 2002

******************************************************

My apologies if you receive this more than once. I hope it is of interest to
you. Also forgive the formality of this email (I’m generating it from a
database).

Due to lots of impending deadlines for everyone, we’re extending the ECOMAS 2002
Deadline to March 11, 2002.
------------------------------------
Deadline Extended to March 11, 2002:

Evolutionary COmputation and Multi-Agent Systems

(ECOMAS 2002)

A Birds-Of-A-Feather Workshop At
GECCO 2002

Description Of The Workshop Topic
=====================================

Multi-agent systems (MAS) are collections of interacting autonomous entities.
The behaviour of the MAS is a result of the repeated asynchronous action and
interaction of the agents. Understanding how to engineer adaptation and
self-organisation is thus central to the application of agents on a large scale.
Moreover, multi-agent simulations can also be used to study emergent behaviour
in real systems.

Desirable self-organisation is observed in many biological, social and physical
systems. However, fostering these conditions in artificial systems proves to be
difficult and offers the potential for undesirable behaviours to emerge. Thus,
it is vital to be able to understand and shape emergent behaviours in agent
based systems. Current mathematical and empirical tools give only a partial
insight into emergent behaviour in large, agent-based societies. EC provides a
paradigm for addressing this need. Moreover, EC techniques are inherently based
on a distributed paradigm (natural evolution), making them particularly well
suited for adaptation in agents.

At the same time, ideas from natural ecosystems or economies, such as resource
flows, niches, and spatial context or neighbourhood can contribute both to the
development of MAS and to the improvement of EC techniques. The interaction
between these different sources of natural inspiration and the two computing
disciplines of MAS and EC is beginning to stimulate a range of systems with
properties that extend the MAS and EC concepts in new and interesting
directions.
Notable examples of systems of that begin to examine the issues of EC in MAS
include Holland's ECHO system, Tierra, Avalanche, Egglets, Amalthaea,
InfoSpiders, and many others.


The workshop follows ECoMAS 2001, which was conducted at GECCO-2001. With more
than 70 attendees, ECoMAS 2001 was, in our opinion, a great success. As a result
of that workshop (and in order to address a pressing need shared by the majority
of the attendees), we have created, organised and launched a new Internet
interested community: the ECoMAS Community.


Community homepage: www.csm.uwe.ac.uk/~rsmith/ECOMAS/index.htm


There is also a discussion forum associated to the community:


Forum homepage: http://groups.yahoo.com/group/ecomas


The goal of the workshop is to maintain a dialog among researchers and
practitioners who are examining EC in MAS. The workshop represents an important
opportunity for those active or interested in this area to hear about current
work, discuss future directions and priorities, and form invaluable research
contacts. We also see the workshop as the natural location for reporting and
enhancing ECoMAS Community activities.

Interest To The GEC Community
=====================================

With the advance of computational power and communications speed, we now live in
a computational world where a large number of agents may be working on behalf of
any given user. A large number of Internet software agents may be acting on
behalf of even the most casual user: searching for music, comparing pension
schemes, purchasing good and services, identifying chat partners, etc. Moreover,
these agents may be collaborating with those of other users, while spawning and
managing agents of their own. In more formal settings, a business, academic, or
government user may simultaneously employ many software agents to manage
workflow, trade goods or information, collaboratively solve problems, etc. In
the future, even relatively simple household appliances may play a role in this
churning system of interacting, computational agents.


In this world, EC theories and practices have new implications. Agents that
interact according to these theories are no longer locked inside the laboratory
conditions imposed by EC researchers and users. The interest in merging the EC
and MAS research communities is certainly growing. In the opinion of the
organizers, it is important to the GEC community that there is a forum to
discuss the particular issues of EC in MAS. Simultaneously, such a forum allows
ideas from contemporary MAS research to spread to the GEC community, providing
the community itself with a chance to address the need of EC embodiment in a
real environment.


Accordingly, the organizers see GECCO 2002 as the natural location to hold the
ECoMAS 2002 workshop.

Workshop Format
=====================================

In the opinion of the organizers, it is important that a workshop involve more
than talks and presentations. Therefore, the workshop will be focused on an
extensive, directed discussion on the future of EC in MAS. Other aspects of the
workshop will be directed at facilitating this discussion:


1)     The workshop will allow the selected presenters to post “mini-posters.”
Much of this material will be available before the workshop, via this web site.


2)     The first segment of the workshop will consist of “mini-presentations” to
preview the mini-poster session. Authors will be allowed to present a strictly
limited number of transparencies. Time constraints will be adjusted, depending
on the number of presenters selected, but a limit will be maintained, to allow
for the sessions outlined below.


3)     The second segment of the workshop will consist of a mini-poster session.


4)     The third segment of the workshop will focus on a discussion of the
future of EC in MAS.


5)     The final, and perhaps most important, segment of the workshop will be a
discussion focused on action items for advancing EC in MAS. The organizers feel
that explicitly providing time to discuss agendas in the fashion will give the
workshop an atypical, meaningful outcome.

Submission Instructions
=====================================
DEADLINE EXTENDED!!!
If you would like to present material at the workshop please submit a 4 page
extended abstract in Postscript or PDF form to cefn.hoile@... by March 11th,
2002. If your submission is accepted, expect to submit a camera ready version of
the extended abstract by April 23rd, 2002, and to submit a web-based
presentation (PowerPoint, HTML, PDF, etc.) by June 1st, 2002.
If you would like to participate, but not present, please notify
cefn.hoile@...  by March 11th, 2002, as GECCO requires us to submit a
participants list.

Important Dates:
=====================================

Submissions Due: March 11, 2002

Review Decisions To Authors: March 25, 2002

Camera Ready Due: April 23, 2002

Web Materials Due: June 1, 2002

GECCO 2002 Dates: July 9-13, 2002


Workshop Organizers
=====================================



Claudio Bonacina, Robert Smith
Intelligent Computing Systems Centre
University of The West of England
Coldharbour Lane, Frenchay
Bristol BS16 1QY, UK




Cefn Hoile, Paul Marrow
Intelligent Systems Laboratory, BTexaCT
Admin 2 PP 5, Adastral Park
Ipswich IP5 3RE, UK

-----Original Message-----
From: Shivanajay Marwaha [mailto:shivanajaym@...]
Sent: 30 January 2002 02:02
To: ecomas@yahoogroups.com
Subject: RE: [ecomas] anyone working on Ants/swarm (Mobile agents)

Hi,

I am working on a new Ant based routing algorithm for computer networks. Ants can distribute the network information (topology, congestion, link banwidth etc.) & reduce the overhead of exchanging routing table updates. (My application requires least communication overhead.) Plus the ant (mobile) code can be minimized by storing the same at the nodes & controlling the time to migrate of ants, in effect controlling the ant traffic.

Are you also employing Ant algorithms for Comp network routing application.

Regards,

Shivanajay

Shivanajay,

We (the DIET project in BTexact's Future Technologies Group) are working to
exploit the dynamics of large populations of lightweight and reactive mobile
agents.

If you can, please post a brief description of your project to the list for
everyone's interest. Look forward to hearing from you.

Re: Conferences, below are some suggestions. This is not intended to be
comprehensive. I hope others on the list will make their suggestions, and
perhaps we can get a list compiled.

Good luck

Cefn
ECOMAS Team


Of course, you should consider GECCO
http://www-illigal.ge.uiuc.edu:8080/GECCO-2002/index.html and CEC
http://www.wcci2002.org/cec/call.html (the two conferences with which my
team are closely associated - the ECOMAS workshop was hosted at GECCO last
year, and we have been invited back again this year).

These are some links drawn from EVONET. Specifically
http://evonet.dcs.napier.ac.uk/evoweb/site_tools/keywordlist5.10.html

ANTS 2002 From Ant Colonies to Artificial Ants: Third International Workshop
on Ant Algorithms
http://evonet.dcs.napier.ac.uk/evoweb/news_events/conferences/conf736.html

Brussels, Belgium, 11 - 14 September 2002  AAAA ALife, Adaptive Behavior,
Agents and Ant Colony Optimization Special Track @ GECCO-2002
http://evonet.dcs.napier.ac.uk/evoweb/news_events/conferences/conf708.html

New York City, NY, USA, 9 - 13 July 2002 GWAL-5 5th German Workshop on
Artificial Life
http://evonet.dcs.napier.ac.uk/evoweb/news_events/conferences/conf705.html

Lübeck, Germany, 18 - 20 March 2002 BioSP3 Workshop on Bio-Inspired
Solutions to Parallel Processing Problems
http://evonet.dcs.napier.ac.uk/evoweb/news_events/conferences/conf663.html

New York City, New York, USA, 9 - 13 July 2002  Operational Research Society
(UK): Local Search Study Group
http://evonet.dcs.napier.ac.uk/evoweb/news_events/conferences/conf639.html

City University, London, UK, 16 - 17 April 2002 PPSN VII The 7th
International Conference on Parallel Problem Solving from Nature
http://evonet.dcs.napier.ac.uk/evoweb/news_events/conferences/conf613.html



These are some of the conference links which came up in my mailbox re:
mobile agents and may be relevant...


The First International Joint Conference on Autonomous Agents & Multi-Agent
Systems AAMAS 2002 Palazzo Re Enzo, Bologna, Italy July 15-19, 2002
http://lia.deis.unibo.it/aamas2002  OR http://www.autonomousagents.org/2002/
[We are presenting two papers from our team here. Further papers have been
accepted from other members of the DIET consortium]

Second International Workshop on ``Agent Based Cluster and Grid Computing''
http://www.cs.cf.ac.uk/User/O.F.Rana/agent-grid-2002/

Sixth International Workshop CIA-2002 on COOPERATIVE INFORMATION AGENTS,
September 18 - 20, 2002, Madrid, Spain
http://www.dfki.de/~klusch/cia2002.html

Coordination and Component-Oriented Computing (Languages, Models, Systems)
http://www.cs.fit.edu/~rmenezes/pdpta02/

17th ACM Symposium on Applied Computing (SAC 2002) Special Track on Agents,
Interactions, Mobility, and Systems (AIMS) AND Special Track on Coordination
Models, Languages and Applications March 10-14, 2002 Madrid, SPAIN
http://www.acm.org/conferences/sac/sac2002/

International Workshop on Software Engineering for Large-Scale Multi-Agent
Systems (SELMAS 2002) Buenos Aires, Argentina May 19, 2002 In conjunction
with the ICSE 2002 (Int'l Conference on Software Engineering)
http://www.teccomm.les.inf.puc-rio.br/selmas2002/

IDEAL 2002 http://ideal02.ee.umist.ac.uk


These conferences have happened, but may provide useful links...

European Conference on Artificial Life http://www.cs.cas.cz/ecal2001/

Second International Workshop on Electronic Commerce (WELCOM'01) Mobility
and Electronic Commerce http://www.informatik.tu-darmstadt.de/GK/welcom/

5th IEEE INTERNATIONAL CONFERENCE ON MOBILE AGENTS December 2-4, 2001
Atlanta, Georgia, USA http://www.cs.dartmouth.edu/MA2001/

Fifth International Workshop CIA 2001 on COOPERATIVE INFORMATION AGENTS
September 6 - 9, 2001 Modena, Italy http://www.dfki.de/~klusch/cia2001.html

AGENT THEORIES, ARCHITECTURES, AND LANGUAGES (ATAL-2001) Seattle, USA ---
August 1-3, 2001 (Immediately before IJCAI-2001) http://www.atal.org/

5th International Conference on Autonomous Agents
http://www.csc.liv.ac.uk/~agents2001/

-----Original Message-----
From: Shivanajay Marwaha [mailto:shivanajaym@...]
Sent: 29 January 2002 13:58
To: ecomas@yahoogroups.com
Subject: [ecomas] anyone working on Ants/swarm (Mobile agents)


Hi,
Is anyone working on mobile agents (ants or swarm intelligence). Is there
any conference/journal on the same in the near future.
Kindly send any CFP in this field.
Thanks,
Shivanajay

---
ECOMAS Web site: http://www.csm.uwe.ac.uk/~rsmith/ECOMAS/
Discussion Group Site: http://groups.yahoo.com/group/ecomas


Your use of Yahoo! Groups is subject to the Yahoo! Terms of Service.

Hi,

Is anyone working on mobile agents (ants or swarm intelligence). Is there any conference/journal on the same in the near future.

Kindly send any CFP in this field.

Thanks,

Shivanajay

Hello K.

Now you are registered, you can go to the files at...

http://groups.yahoo.com/group/ecomas/files/

...where you will find the list of Agent Toolkits.

Cefn
ECOMAS Team

-----Original Message-----
From: ramank0 [mailto:ramank0@...]
Sent: 25 January 2002 23:25
To: ecomas@yahoogroups.com
Subject: [ecomas] Re: Agent Building????


From: K. Raman

I have just joined this discussion group.
I am interested to know about the list of Agent building toolkits.

Could you please let me know Where I can find this list of toolkits.

Thanks.
K. Raman
ramank0@...



--- In ecomas@y..., cefn.hoile@b... wrote:
> I have uploaded a list of Agent toolkits which we compiled earlier
this
> year. Hope this is of some use, Shivanajay. Do any ECOMAS members
have
> experience with these toolkits?
>
> Would be good to hear comments if you feel strongly about the
> excellence/awfulness of any of these.
>
> Also if you have experience/opinions regarding the appropriateness
of any of
> these toolkits to different applications, please let me know. Would
be good
> to improve this resource.
>
> Cefn
> ECOMAS ORGANISER
>
> -----Original Message-----
> From: Shivanajay Marwaha [mailto:shivanajaym@y...]
> Sent: 06 October 2001 07:12
> To: ecomas@y...
> Subject: [ecomas] Agent Building????
>
>
> Hi all,
> Wanted to ask something.
> What is the prefered way of making MAS.
> I mean using some Agent builder software like Zeus



To unsubscribe from this group, send an email to:
ecomas-unsubscribe@yahoogroups.com
---
ECOMAS Web site: http://www.csm.uwe.ac.uk/~rsmith/ECOMAS/
Discussion Group Site: http://groups.yahoo.com/group/ecomas


Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/

From: K. Raman

I have just joined this discussion group.
I am interested to know about the list of Agent building toolkits.

Could you please let me know Where I can find this list of toolkits.

Thanks.
K. Raman
ramank0@...



--- In ecomas@y..., cefn.hoile@b... wrote:
> I have uploaded a list of Agent toolkits which we compiled earlier
this
> year. Hope this is of some use, Shivanajay. Do any ECOMAS members
have
> experience with these toolkits?
>
> Would be good to hear comments if you feel strongly about the
> excellence/awfulness of any of these.
>
> Also if you have experience/opinions regarding the appropriateness
of any of
> these toolkits to different applications, please let me know. Would
be good
> to improve this resource.
>
> Cefn
> ECOMAS ORGANISER
>
> -----Original Message-----
> From: Shivanajay Marwaha [mailto:shivanajaym@y...]
> Sent: 06 October 2001 07:12
> To: ecomas@y...
> Subject: [ecomas] Agent Building????
>
>
> Hi all,
> Wanted to ask something.
> What is the prefered way of making MAS.
> I mean using some Agent builder software like Zeus

Mik wrote: "You also have a multitude of systems composed of more than one
fish.  Normally they act as a cohesive whole (probably with less
than dimensionality x as there are things an individual fish can
do that a school cannot).  But they also give the individuals the
opportunity to do things with other fish that are simply not
possible for a single fish, which would argue that the dimensionality
could also be higher than x*n."

CEFN: If the individual fish have a state space of dimensionality x, I think
it is impossible for the sum of states of n fish to have a higher
dimensionality, since they will all reduce to some function of the space
x*n. This is the case unless you add some extra elements to the system,
increasing its dimensionality.

So, for example, two point fish are moving in three dimensional space and
hence the system's instantaneous state is reduced to x*n = 6*2 (position and
motion in space for two points).

You could propose a new degree of freedom - degree of following, which is
the projection of one fish's normalised motion vector onto the other's
normalised motion vector. This is a higher-level property, which indicates
the degree to which one fish is adopting the same direction as another.
However, this does not mean it is an additional degree of freedom, since it
can actually be fleshed out in terms of the space x*n, which details each
fish's position and vector.

What can happen is that the space containing the states accessible (given an
initial state, and a set of dynamic equations - i.e. the attractor) has a
fraction of the dimensions of the larger state. I think this is what Rob is
getting at, and it is an interesting way of looking at emergence, since it
seems natural that degrees of freedom would be lost through the coupling of
subsystems.

I will throw a definition of emergent out there, and see if the 'cat licks
it up'.

"is Emergent : True of a system where microstates are divergent, and
macrostates are convergent".

Thus, where the fate of individual cells in a body, (i.e. becoming a kidney
cell or not), may be very sensitive to initial conditions, it is rare that
nothing becomes a kidney cell. In fact the process converges quite
specifically on an end state in which kidney, heart, spine, skin etc. etc.
are all represented. The lower level units from which microstates can be
defined are divergent, whilst the macrostate (of something being a kidney
cell) is convergent.

This seems to work for many of the interesting cases of emergence. I would
be interested in counterexamples.

Regards,


Cefn Hoile
ECOMAS Team

Claudio Bonacina wrote:
>
> What if we use the concept of "objective" to define and distinguish
> between SO and emergence?

Well, call it benefit rather than objective and I'll go along with
that for now.  Objective implies to me that there's some planning
and thought (conscious decision making) gone into it.  Evolutionary
systems tend to arrive at the most beneficial behaviour through
semi-random experimentation.

> I think we can talk of self-organization if the system is going
> toward the maximization/minimization of a function (the survival of
> the members of the school, in the school of fish) and this is the
> (one of the) objective of the system.

Yes, but again try to avoid imparting a sense of planning.  With
a school of fish we've got a set of behaviours that prove
beneficial to the fish.  Very probably it is the optimal set of
behaviours for them in that enviornment with those preditors,
so yes, they are min/maxing things.

> The reduction of the space dimensions or of the entropy of the
> system could be seen as a side effect (or perhaps as the objectives
> of the system, although, for example, I'm not sure the entropy of
> the system "school of fish" is decreasing when a predator arrives).

If memory serves they divide and spread, then reform. This makes
the situation the preditor has to deal with more complex, often enough
to confuse him into going hungry.

> With regard to emergence, I think I can agree with Mik: "Emergence
> is basically the ability to connect simple things together so that
> you get more than the sum of the parts." Although I don't think that
> the parts of the system "must" be simple.

Agreed, just simpler than the combined system.

> Here there is no notion of objective. In my opinion the sand cones
> are an emergent property of the (passive) interactions among the
> sand grains and between those and the environment (gravity, impact
> surface, etc.)

Yes, simply sand grains being sand grains in a gravitational
field.  No conscious objective, just behaviour.  Being and doing.
Only by the time you get to complex biological systems do we
start to see any evidence of conscious objective forming and
planning.  Conscious self-organization is of a different nature
to 'natural' self organization. (Hmmm.  Are the fish conscious or
somewhere in between?  Behaviour from instincts would probably
be natural, while learnt behaviour would be conscious (or
programmed).)

> In my opinion there is no self-organization without emergence and
> when emergence allows the system to move closer to an objective we
> can talk of self-organization.

Hmmm. Is that because we detect self organization through
recognition of the 'emergent' behaviour?  Can something be
self organizing if it doesn't produce any characteristic patterns
of organization as a result?

From: Robert E. Smith [mailto:robert.smith@...]

> In my view, one fish would have a characteristic attractor
> dimensionality of some number x. This number is a characteristic of
> the agent's world, it's rule's of behaviour, etc. The question of
> self organization (as I'm posing it here, and I'm brainstorming, so
> shoot me down at will) is whether when you add together n such fish,
> does their resulting behaviour have dimensionality n*x, or simply x,
> or, most interestingly, a behaviour of dimensionality somewhere
> between these two?

Well, at any point each fish is working out for itself what it
will do next, so you've always got the potential for n*x.

You also have a multitude of systems composed of more than one
fish.  Normally they act as a cohesive whole (probably with less
than dimensionality x as there are things an individual fish can
do that a school cannot).  But they also give the individuals the
opportunity to do things with other fish that are simply not
possible for a single fish, which would argue that the dimensionality
could also be higher than x*n.

Mik

Hi Everybody,

it seems that "self-organization vs. emergence" ii a hot issue...I'll try to
do my bit, at least to keep the discussion going.

What if we use the concept of "objective" to define and distinguish between
SO and emergence?

I think we can talk of self-organization if the system is going toward the
maximization/minimization of a function (the survival of the members of the
school, in the school of fish) and this is the (one of the) objective of the
system.
The reduction of the space dimensions or of the entropy of the system could
be seen as a side effect (or perhaps as the objectives of the system,
although, for example, I'm not sure the entropy of the system "school of
fish" is decreasing when a predator arrives).

With regard to emergence, I think I can agree with Mik: "Emergence is
basically the
ability to connect simple things together so that you get more
than the sum of the parts." Although I don't think that the parts of the
system "must" be simple.
Here there is no notion of objective. In my opinion the sand cones are an
emergent property of the (passive) interactions among the sand grains and
between those and the environment (gravity, impact surface, etc.)

In my opinion there is no self-organization without emergence and when
emergence allows the system to move closer to an objective we can talk of
self-organization.

Perhaps I'm completely out of track but I'm interested in your comments.

Claudio

-----Original Message-----
From: Robert E. Smith [mailto:robert.smith@...]
Sent: 16 December 2001 16:41
To: ecomas@yahoogroups.com
Subject: RE: [ecomas] Introduction and Discussion


Hi Mik,

> Yes, but that sounds fairly hard to measure.  If you pour sand
> out of a bottle, do you consider the nice cone it makes to be
> self organized (or environmentally organized, given that the
> grains of sand are passive entities).  If it isn't self organized,
> what's going around making all the little cones?


Well, are they passive? They do interact, via (local) contact dynamics,
under the influence of gravity. So, yes, I'd say they are self organized,
although not in a way that's terribly interesting.

> Yes, I'd say the behaviour of the school was emergent, emerging
> from the fish changing their organization (moving around) to respond
> to their environment.  With my approach the schooling behaviour is
> definitely missing if there is only one fish.  With yours I'm not
> so sure. Would it be in a school of one?


In my view, one fish would have a characteristic attractor dimensionality of
some number x. This number is a characteristic of the agent's world, it's
rule's of behaviour, etc. The question of self organization (as I'm posing
it here, and I'm brainstorming, so shoot me down at will) is whether when
you add together n such fish, does their resulting behaviour have
dimensionality n*x, or simply x, or, most interestingly, a behaviour of
dimensionality somewhere between these two?

R.








---------------------------------------------
Robert E. Smith
Director
Intelligent Computer Systems Centre
University of The West of  England
for further info and contact details, see:
http://www.csm.uwe.ac.uk/~rsmith
---------------------------------------------



Yahoo! Groups Sponsor
ADVERTISEMENT




To unsubscribe from this group, send an email to:
ecomas-unsubscribe@yahoogroups.com
---
ECOMAS Web site: http://www.csm.uwe.ac.uk/~rsmith/ECOMAS/
Discussion Group Site: http://groups.yahoo.com/group/ecomas


Your use of Yahoo! Groups is subject to the Yahoo! Terms of Service.

I don't have a neat definition of self-organisation, but I'd say it must
involve some form of spontaneous decrease in entropy. Decrease in entropy (I
think) corresponds to a reduction in the state space volume occupied by the
attractor. Presumably a reduction in the dimensionality of the attractor
would involve quite a significant reduction in volume ... so I think I'm in
general agreement with Rob's view of SO.

[BTW Mik, I'm a silent lurker on the NECSI list so am familiar with your
name from there ... didn't realise you worked on IBM heavy metal though!]


	 Paul Kearney
	 Solutions Research Laboratory
	 BTexact Technologies

	 e-mail:paul.3.kearney@...
	 tel: +44(0) 1473 646835

	 Antares (Admin) Building pp2/4, Adastral Park, Martlesham, Ipswich
IP5 3RE, UK

	 BTexact Technologies is a trademark of British Telecommunications
plc
	 Registered office: 81 Newgate Street London EC1A 7AJ
	 Registered in England no. 1800000

	 This electronic message contains information from British
Telecommunications plc which may be privileged or confidential. The
information is intended to be for the use of the individual(s) or entity
named above. If you are not the intended recipient be aware that any
disclosure, copying, distribution or use of the contents of this information
is prohibited. If you have received this electronic message in error, please
notify us by telephone or email (to the numbers or address above)
immediately.




-----Original Message-----
From: Michael Clarke [mailto:mykael@...]
Sent: 14 December 2001 14:40
To: ecomas@yahoogroups.com
Subject: Re: [ecomas] Introduction and Discussion


> "Robert E. Smith" wrote:
>
> Hi Mik,
>  Really interesting stuff. Are you considering submitting an abstract
> to ECOMAS 2002? Although you don't have an EC element, I personally
> think that the real-world nature of what you are doing might make for
> interesting conversation at the workshop.

I'm probably not going to attend as I'm based in Australia, although
I might turn up at the alife conference in Sydney next year.

> WRT Emergence and Self-Organization, I don't believe there are any
> firm definitions.

Now that I'll agree with you on.

> but the working ones I've been using lately *might*
> be summarized as follows.
>
> Self-organization in a system of n interconnected agents relates to an
> attractor in some phase space of the system behaviour whose
> dimensionality is less than n*(the number of dimensions for a single
> agent), and greater than (the number of dimensions for a single
> agent).

Yes, but that sounds fairly hard to measure.  If you pour sand
out of a bottle, do you consider the nice cone it makes to be
self organized (or enviornmentally organized, given that the
grains of sand are passive entities).  If it isn't self organized,
what's going around amking all the little cones?

> Emergence is the capability to make rapid transitions in the
> dimensionality of this attractor under small changes in internal or
> environmental parameters.

Hmmm. That sounds more like chaotic behaviour - a small change in
the input can give a big change in the output.  I generally
think that the energence has to be visible at a higher level than
the level that the agents work at.

> These ideas are suggested by the self-organization and emergence in a
> school of fish (for instance), where they behave as a dynamic unit
> (but not as a single entity), and this behaviour emerges from and
> responds to changes in environmental parameters (e.g., the entrance of
> a predator into the school).
>
> But, in these working definitions (unlike yours), you can have
> self-organization without emergence, but not the other way around!
>
> In fact, I think you and I have just transposed the meanings of these
> two terms.

Yes, I'd say the behaviour of the school was emergent, emerging
from the fish changing their organization (moving around) to respond
to their enviorment.  With my approach the schooling behaviour is
definetly missing if there is only one fish.  With yours I'm not
so sure. Would it be in a school of one?

Mik
Mik


To unsubscribe from this group, send an email to:
ecomas-unsubscribe@yahoogroups.com
---
ECOMAS Web site: http://www.csm.uwe.ac.uk/~rsmith/ECOMAS/
Discussion Group Site: http://groups.yahoo.com/group/ecomas


Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/

>>>
Interesting.  My initial response is that those selfish agents
are going to cause you problems.  Because they are each self
optimizing for their task, you are very likely to hit problems
due to micro-optimization (where each individual is running
very efficiently, but the overall system throughput sucks).
>>>

Not at all. In fact, how to ensure that doesn't happen is part of what
our math gets you. We now have about a dozen published papers
demonstrating this in various domains. (Most are provided on our web
site.)

--
David Wolpert

Automated Learning Group
NASA Ames Research Center
MS 269-1, Moffett Field, CA 94035, USA
dhw@...
650-604-3362 (v)
650-604-3594 (f)
http://ic.arc.nasa.gov/~dhw/




"Confusion is a precondition for insight."


"Wipe your hand across your mouth, and laugh.
The worlds revolve like ancient women,
gathering fuel in vacant lots."

> "Robert E. Smith" wrote:
>
> Hi Mik,
>  Really interesting stuff. Are you considering submitting an abstract
> to ECOMAS 2002? Although you don't have an EC element, I personally
> think that the real-world nature of what you are doing might make for
> interesting conversation at the workshop.

I'm probably not going to attend as I'm based in Australia, although
I might turn up at the alife conference in Sydney next year.

> WRT Emergence and Self-Organization, I don't believe there are any
> firm definitions.

Now that I'll agree with you on.

> but the working ones I've been using lately *might*
> be summarized as follows.
>
> Self-organization in a system of n interconnected agents relates to an
> attractor in some phase space of the system behaviour whose
> dimensionality is less than n*(the number of dimensions for a single
> agent), and greater than (the number of dimensions for a single
> agent).

Yes, but that sounds fairly hard to measure.  If you pour sand
out of a bottle, do you consider the nice cone it makes to be
self organized (or enviornmentally organized, given that the
grains of sand are passive entities).  If it isn't self organized,
what's going around amking all the little cones?

> Emergence is the capability to make rapid transitions in the
> dimensionality of this attractor under small changes in internal or
> environmental parameters.

Hmmm. That sounds more like chaotic behaviour - a small change in
the input can give a big change in the output.  I generally
think that the energence has to be visible at a higher level than
the level that the agents work at.

> These ideas are suggested by the self-organization and emergence in a
> school of fish (for instance), where they behave as a dynamic unit
> (but not as a single entity), and this behaviour emerges from and
> responds to changes in environmental parameters (e.g., the entrance of
> a predator into the school).
>
> But, in these working definitions (unlike yours), you can have
> self-organization without emergence, but not the other way around!
>
> In fact, I think you and I have just transposed the meanings of these
> two terms.

Yes, I'd say the behaviour of the school was emergent, emerging
from the fish changing their organization (moving around) to respond
to their enviorment.  With my approach the schooling behaviour is
definetly missing if there is only one fish.  With yours I'm not
so sure. Would it be in a school of one?

Mik
Mik

David Wolpert wrote:
>
> Hi,
>
> Here's an overview of the work my group's doing for the past several
> years. We're also co-organizing a Santa Fe Institute workshop, a
> spring AAAI 02 symposium, a WCCI 02 special session, and possibly a
> NASA workshop on these themes.

Interesting.  My initial response is that those selfish agents
are going to cause you problems.  Because they are each self
optimizing for their task, you are very likely to hit problems
due to micro-optimization (where each individual is running
very efficiently, but the overall system throughput sucks).

On the necsi complexity list we had a case where a guy was able
to increase packet throughput quite dramatically by adding
an element into the routing algorythm that deliberatly
chose an apparently sub-optimal solution.  This increased overall
bandwidth utilization and reduced the queues on the links that
the traditional algorytems said were the fastest ones.

The other problem will be in how you detect and reward altristic
behaviour.  In my networks we have rules that make the agents
calculate data values that they never use (like 'are all agents
downstream from me offline'). Why? Well somewhere upstream
from it there may be a resource where this knowledge is important.
It doesn't benefit the individual node, but it is essential to
the overall functioning of the system in some circumstances.

Consider that the objective may well change over time.  Any ideas
about how to allow the agent to retain the behaviours necessary to
satisfy any one of three or four different objectives?  The point
here is not to optimize away behaviours that are only significant
for the achievement of some of the objectives.

Mik

Norberto Eiji Nawa wrote:

> Mik> With the agent systems, I'm interested in seeing if it's possible
> Mik> to get them to learn without making mistakes.  In the
> Mik> systems mangement field, the software is installed to reduce and
> Mik> remove mistakes.  This is why the agent system is so tightly
> [cut]
>
> What specific learning algorithm/architecture do you have in mind for
> the agents?

I don't know.  Most of the ones I've seen are esentially
genetic systems - try something at random, make an artifical
fitness evaluation and cull the bad ones.  Not really
suitable for what I'm trying to do.  I suppose some sort of
scenario where the system 'learned' by watching state changes
as operators drove the system and remembered the sequences
might work.

> As you are working with a real system, as opposed to a
> minimum-canonical frameworks usually found in the literature where the
> agents have equal learning capacities and learn synchronously, I think
> there are several interesting issues to be tackled, such as how to
> decide when an agent should update its internal states, etc.

Yeah. We found we had to lock the whole struture to just make a simple
update, because we couldn't predict which resources the update was
going to affect, and it was possible for separate updates to
dead lock.  We also had to put an evaluation sequence for the
behavioural rules in, because the rules themselves formed a
hierarchy - local rules, neighbourhood rules and finally output
rules.  All agents evaluate their local rules first, then they
all evaluate their neighbourhood rules (taking data from
neighbouring agents) and then they all evaluate output rules
that generate actions to be taken.  There are hierarchies of
rules within each of these rule sets.

> Mik> 3. Agent systems can work fine without evolutionary capabilities.
>
> Even though I understand the distinction between evolution and
> learning in this context, wouldn't learning in a multi-agent system be
> as harmful as evolution?

Well, that depends on whether or not we can get them to learn (or
evolve) in a useful fashion.  Trial and error processes are a
no-no because it's a critical system.  But things like the
rule sets could possably be evolved in an offline enviornment before
being used in production.  Likewise there should be some potential
to learn through observation of a correctly operated system.

Mik

Mik,

thanks for you very interesting post. Let us know more about your work!

Mik> With the agent systems, I'm interested in seeing if it's possible
Mik> to get them to learn without making mistakes.  In the
Mik> systems mangement field, the software is installed to reduce and
Mik> remove mistakes.  This is why the agent system is so tightly
[cut]

What specific learning algorithm/architecture do you have in mind for
the agents? As you are working with a real system, as opposed to a
minimum-canonical frameworks usually found in the literature where the
agents have equal learning capacities and learn synchronously, I think
there are several interesting issues to be tackled, such as how to
decide when an agent should update its internal states, etc.

Mik> 3. Agent systems can work fine without evolutionary capabilities.

Even though I understand the distinction between evolution and
learning in this context, wouldn't learning in a multi-agent system be
as harmful as evolution?

Looking forward for some (many) opinions,

Eiji